Saturday, February 20, 2016

The Moore the Merrier?

Hello chip and microprocessor fans – and I know you are – I’m sure you’ve heard of Moore’s Law, invented in 1965 by Intel co-founder Gordon Moore. Today, he’s old (87) with an estimated net worth of $6.7 billion. The above photo was taken 14 years ago!
In 1965, Gordon E. Moore was working as the director of research and development… at Fairchild Semiconductor. He was asked by Electronics Magazine to predict what was going to happen in the semiconductor components industry over the next ten years. In an article published on April 19, 1965, Moore observed that the number of components (transistors, resistors, diodes or capacitors) in a dense integrated circuit had doubled approximately every year, and speculated that it would continue to do so for at least the next ten years. In 1975, he revised the forecast rate to approximately every two years. Carver Mead [Cal Tech professor] popularized the phrase ‘Moore's law.’ The prediction has become a target for miniaturization in the semiconductor industry, and has had widespread impact in many areas of technological change.” Wikipedia.
Was this prediction a physical law or a self-fulfilling prophecy? After all, Moore kind of guided Intel and helped set goals. Can we continue to keep doubling our processing speed at the same rate or are we hitting a wall?
“For the better part of the '70s, '80s, '90s, and into the first decade of the 21st century, Moore’s law proved to be correct—but only because it was a self-fulfilling prophecy, according to a recent article in the science journal Nature[the International Weekly Journal of Science], which argues that the prophecy is about to come to an end.
“Instead of being a natural consequence of improved engineering and fabrication methods, the amount of transistors on a chip only doubled every few years-–as Moore’s law said it would—because chipmakers deliberately chose to match what the law said by agreeing to produce microprocessors on a roadmap that synced with Moore’s prediction.” FastCompany.com, February 10th. OK, let’s buy into that theory for a moment. What are we seeing in the physics of microchip and microprocessor development? It does seem that this “doubling” assumption is indeed hitting that wall.
Next month, says Nature, the worldwide semiconductor industry will formally acknowledge it can no longer keep pace with the doubling of transistors every two years… The reasons are due to physics and economics. As microprocessors circuits have shrunk—currently about 14 nanometers across—the increasing the number of circuits packed into a smaller size leads to greater amounts of heat given off. When devices used to be in the form of large desktops, this heat was easier to dissipate. But our go-to computing devices are now smartphones, and their small size doesn’t allow for the kind of heat dissipation that future, increasingly faster microprocessors would require. As Nature points out, no one wants to hold a hot smartphone.
“The other reason for abandoning the Moore’s law roadmap is economic. In order for silicon chips to do more, they must pack more transistors in a smaller space. The smaller the chips become, and the more transistors they pack in, the more costly they are to manufacture, due to new fabrication facilities needed to produce them. Each fab facility can cost billions of dollars, and not many companies have the financial resources to support them.” FastCompany.
We’re already processing data very quickly, and in the mainframe world, between grid computing and developments in plasma computing, speed is accelerating very fast. We are going keep growing, speeding up, discovering new physical capacities and develop the kind of software that will use it efficiently. Artificial intelligence is not slowing down either.
I’m Peter Dekom, and while this won’t impact most of us in quantum leaps, it does suggest that we need to adjust our expectations to a more tenable reality.

No comments: